Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
1
Mandarin Adaptations of Coda Nasals in English
Loanwords∗
Feng-fan Hsieh, Michael Kenstowicz and Xiaomin Mou
Massachusetts Institute of Technology
The paper documents and analyzes the ways in which English loanwords
into Mandarin are adapted to conform to the Rhyme Harmony constraint
that requires the front vs. back quality of a nonhigh vowel to agree with
the coronal vs. dorsal character of a nasal coda. The principal finding is
that the backness of the English vowel determines the outcome and can
force a change in the place of articulation of the nasal coda. This is
attributed to the phonetic salience of the vowel feature in comparison to
the relative weakness of the nasal place feature. It is concluded that
*An earlier version of this paper was read at the Theoretical East Asian Linguistics Conference held at Harvard University, August 2005. We thank the audience as well as Andrea Calabrese, François Dell, San Duanmu, and Moira Yip for helpful comments.
2
phonetic salience is a critical factor in loanword adaptation that can
override a phonologically contrastive feature.
Keywords enhancement, salience, phonetics vs. phonology
1. Background and Motivation
In the recent theoretical literature on loanword phonology two competing
models have emerged. The first, championed by Paradis & LaCharité
(1997, 2005) and others, holds that loanword adaptation is executed
primarily by bilinguals who draw on their phonological competences in
both the donor (L2) and recipient (L1) languages to discern segmental
equivalences at an abstract, phonological (phonemic) level. When an
exact phonemic match is not found then the closest available phoneme is
chosen, with distance measured in terms of the distinctive features
operative in the native, L1 grammar. An alternative view, typically
couched within the OT model, sees loanword adaptation as based on the
phonetic output of the donor language--either in the form of a raw
acoustic signal (Silverman 1992) or more usually in a UG-based phonetic
3
transcription of varying degrees of detail and abstraction.1 The adapter
can take a variety of factors into account in order to make the loan sound
like a word of the native language while still remaining as faithful as
possible to the source of the loan. These include orthography as well as
phonetic properties that are salient to an L1 speaker--regardless of their
contrastive status in the L1 or L2 grammars. See Kenstowicz & Suchato
(2006) and Yip (2006) as well as cited references for discussion of this
alternative.
Mandarin Chinese presents us with the possibility of an interesting test
of these two alternative models of loanword adaptation. According to
most analyses (e.g. Duanmu 2000, 2007), Mandarin has five vowel
phonemes: /i/, /y/, /u/, /ə/, and /a/. The high vowels contrast for
[back] and [round] while the mid and low vowels do not. Stressed
syllables are subject to a bimoraic constraint. There are no complex
syllable margins. Codas are restricted to the nasals /n/ and /ŋ/ (modulo
r-suffixation in the formation of the diminutive) and the
1 Under standard conceptions, OT grammars lack an intermediate, phonemic (word-level) level of representation, making the kind of mapping envisioned by Paradis & LaCharité (1997, 2005) unavailable.
4
glides/semivowels /j/, /w/. The canonical lexical item has the shape
C(Gl)VX (X = V, Gl, N). The vowels take on a variety of allophonic guises
depending on the surrounding consonants. In (1) we illustrate several
generic CVV syllables. The first column is the Pinyin transliteration, the
second is the underlying phonemicization, and the third is a broad
phonetic transcription (Duanmu 2000).
(1) Pinyin UR PR
ta tha thaa 'she'
ti thi thjii 'dam'
tu thu thwuu 'mud'
te thə thɤɤ 'special'
In the context of nasal codas the low vowel takes a relatively front
allophone before the dental nasal (typically transcribed as [an]) and a
relatively back, unrounded allophone before the velar nasal (transcribed
as [ɑŋ])--a distribution termed Rhyme Harmony in Duanmu (2000, 2007).
5
By contrast, in English front and back low vowels freely combine with the
dental and velar nasal phonemes to give four possible combinations.2
(2) English Mandarin
[æn] Dan [an] dan 'egg'
[æŋ] dang
[ɑn] Don [ɑŋ] dɑŋ 'swing'
[ɑŋ] dong
If loanword adaptation abstracts away from the phonetic details in
both L1 and L2 grammars, then we expect that in cases of conflict
between faithfulness to the English vowel or to the nasal coda, the
Mandarin adaptation should be determined by the nasal consonant. This
is because the nasal coda is the only point of similarity at the
phonological level, given that the vowel is unspecified or noncontrastive
for [back] in Mandarin (indicated by the archiphoneme A; see Wang
2 Before the velar nasal the vowel is rounded [ɐ] or [ɔ] for many English speakers.
6
(1993) and Duanmu (2000, 2007) for details). This scenario is sketched in
(3)
(3) phonological mapping
English Mandarin æn An æŋ ɑn Aŋ ɑŋ
Alternatively, if the adapter is trying to achieve the best phonetic match
then in cases of conflict (i.e. English [æŋ] and [ɑn]), additional
considerations may come into play to decide the outcome. A priori we
might expect variation across different lexical items depending on
whether the vowel or the coda nasal is the determining factor.
Alternatively, the adapter might call on other criteria to break the tie. For
example, while the [±back] vowel difference is phonologically
predictable, it is more salient phonetically and hence could provide a
better overall match than the nasal coda consonant--a segment whose
place features are relatively faint and highly susceptible to neutralization
7
cross-linguistically. The latter scenario predicts the correspondences in
(4).
(4) phonetic (auditory) mapping
English Mandarin æn an æŋ ɑn ɑŋ ɑŋ
In the absence of a well-developed theory of loanword adaptation, it is
unclear which of these two alternatives is more likely to be true. Hence
the empirical study of such conflicting cases is an important step towards
such a theory.
Whether (3) or (4) is the correct scenario turns out to be a question
that is not so easily answered. It is well known that in comparison to
Japanese and Korean, Mandarin Chinese is highly resistant to
phonological loans, preferring loan translations or calques (Novotná
1967). Furthermore, it appears that many of the phonological loans that
entered the language in the Early Modern period (c. 1900 - 1940) have
been become obsolete or been replaced. Contemporary Mandarin
8
vocabulary thus lacks a substantial body of loanwords that we can easily
consult in order to answer our question. We are thus forced to fall back
on more meager resources. We are aware of two sources with relevant
data. First, there is the Dictionary of Loanwords and Hybrid Words in
Chinese (Liu et al. 1984).3 We analyze material drawn from this source in
section 2. Second, there is the Website of the Chinese Ministry of Foreign
Affairs, which has a listing of the preferred transcriptions and
pronunciations for many foreign place names. We analyze data drawn
from the latter in section 3. Section 4 reviews the phonetic basis of the
front-back vowel enhancement of the coda nasal contrast to provide
independent support for our analysis. Section 5 is a brief summary and
conclusion.
2. Analysis of Loanwords from the Dictionary
Our study's loanword corpus consists of c. 600 items drawn from Liu et al.
(1984) that contain a VN sequence in the loan source (typically English). 3 The authors state that the dictionary was constructed in the period 1960-64 from material in dictionaries, monographs, Chinese translations of foreign classics, academic journals, newspapers, magazines, as well as other sources such as import-export catalogs, customs declaration forms, etc. The dictionary contains c. 10,000 items.
9
The discussion here focuses on items where the vowel of the source word
is low or mid since this is where the vowel is phonologically unspecified
or noncontrastive for front vs. back in Mandarin and the resolution of the
conflict between faithfulness to the vowel vs. faithfulness to the coda
nasal of the English word can be studied. We organize the data into
several subcategories. The first consists of English VN rimes where V is
nonhigh, N is a dental or velar nasal, and the syllable bears some degree
of stress. Our main finding is that it is the front vs. back category of the
vowel that determines the outcome. We then look at VN sequences
drawn from final unstressed syllables in English. Here we find
competition between strategies based on approximation to the English
reduced vowel vs. those based on the orthographic representation. The
next category consists of loans in which a nasal has been inserted into the
coda to achieve a bimoraic syllable. Our data indicate that the front vs.
back quality of the vowel in English determines the substitution as [n] or
[ŋ], respectively. In the last group, the coda nasal of the English loanword
is [m], which must be repaired--typically by changing the [m] to [n] or
[ŋ]. Once again, we find that the vowel of the source word decides the
outcome.
10
2.1 VN
When the vowel is low there is a partial correlation between its front vs.
back status in the English source word and its orthographic form. Front
vowel [æ] (RP [a]) is represented with the letter a (e.g. hat) while the
back, unrounded [ɑ] (or rounded [ɒ] in RP) is typically represented with
o (e.g. hot). However, sometimes [ɑ] is also spelled with the letter a (e.g.
class). Since we cannot always rely on English spelling, we have checked
all examples with the OED.4
When the English source consists of a front vowel combined with a
dental nasal ([æn]) or a back vowel combined with a velar nasal ([ɑŋ]),
we expect the Mandarin adaptation to contain a matching rime--i.e. [an]
or [ɑŋ], respectively. For English [æn] there are 31 loans in our corpus;
all but five support this hypothesis.
(5) English Mandarin
a. anchovy [æn] an.chou [an]
4 For the period of the early 20th century the British presence in China was much stronger than the American one and thus British English is the more likely source for the English loanwords.
11
angel an.qi.er
antecedent an.ti.xi.deng
flange fa.lan.(-pan)
Vandyke fan.tai.ke
van de graaf fan.de.ge.la.fu
furan fu.ran
candelilla kan.te.li.la
clan ke.lan
cotangent kou.tan.jin
lancers lan.sa.si
rand lan.te
lanthanium lan
Levant li.fan.de
romantic luo.man.di.ke
romance luo.man.si
mantle man.tuo
pandora pan.duo.la
saraband sa.la.ban.de
sandal shan.da.li
12
Sudan su.dan
tangent tan.jin
b. bandage [æn] beng.dai [ɤŋ]
phalanstery fa.lang.ji.si.te [ɑŋ]
scandium kang
lantum lang.tang
vandal wang.da.er
For English [ɑŋ] rimes there are seven examples in the corpus; five are
adapted as expected with a velar nasal and back vowel allophone (6a).
(6) English Mandarin
a. Congo [ɑŋ] gang.guo [ɑŋ]
franc fa.lang
furlong lang
pingpong ping.pang
mongoose meng.ge [ɤŋ]
b. encore [ɑŋ] an.ge [an]
gong gunge [un]
13
The matrix in (7) summarizes the adaptation of the harmonic rhymes.
Mandarin preserves the front vs. back quality of the rhyme to a significant
degree.
(7) Mandarin
an ɑŋ
æn 26 4
ɑŋ 2 5
English p < 0.008 (two-tailed Fisher's exact test)
In loans where the English vowel and coda nasal do not agree as front
vs. back, there are two ways in which the adaptation can be brought into
alignment with the Mandarin [an] and [ɑŋ] codas required by Rhyme
Harmony. Either the front vs. back character of the vowel can be
preserved and the nasal changed; or alternatively the nasal coda can be
held constant and the vowel adjusted. The data overwhelmingly evidence
the first strategy. The corpus contains 24 loans where the English source
consists of a low, back vowel and a dental nasal. In all of the
14
corresponding Mandarin loans, it is the nasal consonant that is changed,
giving an [ɑŋ] rhyme in the majority of cases (8a). In a few (8b), the
vowel is mid [oŋ] or [ɤŋ].
(8) English Mandarin
a. anon(ym) [ɑn] a.nang [ɑŋ]
ounce ang.si
Browning bai.lang.ning
pound bang
bezant bie.sang
radon dong
Oregon e.le.gang
ergon er.gang
concept gong.si.bu.tuo
fandango (Sp.) fang.dange
geon ji.ang.ding.sheng
condenser kang
canto (It.) kang.tuo
crown ke.lang
15
marathon ma.la.song
monsoon mang.xun
pontificate pang.ti.fei.jia.te
pontoon pang.tong
plante (Fr.) pu.lang.te
samba sang.ba
sonnet shang.lai-ti
b. gondola [ɑn] gong.duo.la [oŋ]
neon ni.hong
cellon se.long
mont meng [ɤŋ]
The number of loanwords with a velar nasal coda and front vowel
nucleus is again smaller (13).5 Only four remain faithful to the nasal (9b).
5 The word mu.si.deng [dɤŋ] < mustang [æŋ] has an unexpected mid rather than low vowel. We are not able to explain this change in height. Given that it is treated as mid, the expected [den] syllable is rare in Modern Mandarin and is avoided in loans. See discussion of mid vowels below. The Yankee > yang.ji loan might arise from semantic contamination since the character used to represent it means 'western'.
16
The rest (9a) change the [ŋ] to [n] so as to remain faithful to the English
vowel.
(9) English Mandarin
a. bank [æŋ] ban.ke [an]
Angora an.gela
Franklin fu.lan
Grange ge.lan.qi
Lancashire lan.kai.xia
Langley lan.le
tango tan.ge
tank tan.ke
triangle te.li.an.ge.'er
b. gangsa [æŋ] gang.sha [ɑŋ]
sarangi sa.lang.ji
wankel wang.ke.er
Yankee yang.ji
17
The matrix in (10) summarizes the resolution of the conflicting English
rimes.
(10) Mandarin
an ɑŋ
æŋ 9 4
ɑn 0 24
English p < 0.000001 (two-tailed)
In sum, our hypothesis is supported--the more salient vowel normally
determines the adaptation even though the nasal coda is the site of the
phonological contrast.
Adopting the approach to loanword phonology taken in Kenstowicz
(2005) and Yip (2006) where faithfulness to the loanword source is
expressed as an OT Output-Output faithfulness constraint that may be
ranked differently from the corresponding Input-Output constraint of
native grammar, we can express the adaptation of the low vowel+nasal
coda words into Mandarin as follows. First, we assume an undominated
markedness constraint of RHYME HARMONY (Duanmu 2000, 2007) that
18
requires a front vs. back low vowel to co-occur with a dental vs. velar
nasal coda, respectively (see Flemming 2003 for discussion of the
phonetic basis for such a constraint). Second, we assume that the nasal
codas are the site of the lexical contrast in Mandarin (F » M) while the
[±back] low vowel allophones [ɑ] and [a] are distributed by Rhyme
Harmony (M » F). Given the OT premise of Richness of the Base, native
grammar inputs in which the nucleus and coda violate Rhyme Harmony
are repaired by faithfulness to the coda, as in (11) below.
(11)
/ɑn/ Rhyme Harmony Id-CPl-Coda Id-[back]
[ɑn] *!
>[an] *
[ɑŋ] *!
/aŋ/
[aŋ] *!
>[ɑŋ] *
[an] *!
19
But in the loanword phonology, in order to be faithful to the vowel of the
source language, the adapter calls on the otherwise submerged Id-[back]
constraint which is “cloned” as an Output-Output constraint between
English and Mandarin and ranked above Faithfulness to CPl-Coda.
(12) Id-[back]E-M » Id-CPl-Coda » Id-[back]
Given this ranking, the input-output adaptation mapping is diverted
towards faithfulness to the otherwise redundant vowel, as shown below.
(13)
/ɑn/ Rhyme Harmony Id-[back]E-M Id-CPl-Coda
[ɑn] *!
[an] *!
>[ɑŋ] *
/æŋ/
[aŋ] *!
[ɑŋ] *!
>[an] *
20
Let us now consider examples where the English source word consists
of a mid vowel followed by a nasal coda. In Mandarin there are four
surface mid vowels whose distribution is determined by the surrounding
onset and coda consonants (Duanmu 2000, 2007). The basic allophone,
found in open syllables, is back unround [ɤ]. As with the low vowel, a
dental nasal requires a more fronted vowel nucleus [ən] (e.g. sen [sən]
‘forest’) while a velar nasal requires a back vowel nucleus [ɤ] or [o]. In
some varieties of Mandarin the latter derives from earlier [uŋ] by
lowering (e.g. [toŋ] ‘same). Dialects also differ in whether or not [oŋ] is
retained after a labial onset: cf. Taiwanese Mandarin [moŋ] 'fierce' vs.
Beijing [mɤŋ]. Finally, there is a more close front vowel allophone after a
palatal glide onset [je].6
6 In order to have some sense of the location of these allophones in phonetic space, we recorded five tokens for each from a male Taiwanese Mandarin speaker (the first author). The results showing the average first and second formant measures and standard deviations from the mid point of the vowel are shown below. We see that the [ə] is a relatively central vowel falling roughly midway between the front [je] and the back rounded [o] in F2. The nucleus of the [je] is more close, showing the influence of the onglide.
21
Turning to the loanwords, there are three cases to consider
depending on whether the English vowel is [ɛ], [o], or [ʌ]. We examine
each of these in turn. First, when the mid vowel is [ɛ] and the coda is [n]
in the English source, Mandarin offers the choice between [je] and [ən].
Neither option is particularly close. It is therefore of some interest that
the former is systematically rejected in favor of the latter (14).
(14) English Mandarin
amen [ɛn] a.men [ən]
pentyl ben.ti.er
benzene ben
benzocaine ben.zuo.ka.yin
Enfield en.fei.er
phen(ol) fen
phosgene fu.su.zhen
F1 F2 ben [bən] 472/9 1476/32 ‘run’ beng [bɤŋ] 471/12 1100/68 ‘collapse’ song [soŋ] 465/20 857/47 ‘loose bjen [bjen] 366/15 2352/31 ‘edge’
22
convention kang.wen.xin
pimento pi.men.ta
cement shui.men.ting
The choice of central [ə] over diphthongal [je] for English [ɛ] indicates
that Dep-Glide dominates faithfulness for [back].
(15)
/bɛn/ Dep-Gl Id-[back]E-M
[bən] *
[bjen] *!
The few exceptions to this correspondence occur when the C[ən] syllable
is either not attested in the existing inventory of Mandarin syllable types
or is rare (16). In this case an adjustment must be made--changing the
vowel or the nasal.7
7Since the data in the dictionary are all transcribed with Chinese characters, a syllable containing a novel combination of CVC cannot be easily represented. It is not clear to us to what extent this fact about
23
(16) Ländler (Germ) lian.de.la *[lən]
lentor lun.tuo *[lən]
engine yin.qing *[ən]
tendency ting.deng.se *[tən], *[tin]
Curiously, it is the vowel height that is changed while the front vs. back
property of the vowel in the source word is largely maintained. This
suggests that faithfulness for the vowel is broken down into faithfulness
for [±back] (F2) vs. faithfulness for height [±high], [±low] (F1), as
indicated in the tableau in (17). (We assume the undominated constraint
USE-LISTED-SYLLABLE: A syllable in the adapted loanword must have a
precedent in the native inventory).
orthography inhibits the creation of novel combinations of onset, rhyme, and tone. See Bauer (1985) for novel syllables in loans to Cantonese.
24
(17)
/ten/ Use-Ld-Syll Id-[back]E-M Id-[high]E-M Id-CPl-Coda
ten *!
tin *!
>tiŋ * *
toŋ *! *
Our corpus contains 12 examples of conflicting English rimes in
[on]. In the corresponding Mandarin adaptations, they change the [n] to
velar [ŋ] in order to remain faithful to the vowel (18a). The lone
exception is shown in (18b). When the onset is a labial consonant then
the vowel [o] is blocked by the labial disharmony constraint that bans the
combination of a labial onset and rhyme in the Beijing dialect. The back
unrounded vowel [ɤ] (Pinyin eng) or the low [ɑ] is substituted in this
case.
(18) English Mandarin
a. amidone [on] a.mi.tong [oŋ]
barbitone ba.bi.tong
25
chalone ka.er.long
clone ke.long
Cologne ke.long
hormone he.er.meng [ɤŋ]
telephone de.lu.feng
microphone mai.ke.feng
sousaphone su.sha.feng
sarrusophone sa.luo.suo.feng
leone li.ang [ɑŋ]
b. scone [on] shi.gan [an]
The contingency table in (19) summarizes the outcome of the
competing changes for the [±back] feature of the vowel and the
corresponding coronal vs. dorsal place feature of the nasal coda.
(19) Mandarin
Vn Vŋ
en 14 2
on 1 11
English p = 0.000053 (two-tailed)
26
As the tableau in (20) shows, in the case of the conflicting back vowel +
coronal coda the correct adaptation is made by the Id-[back]E-M » Id-CPl-
Coda ranking established for the low vowels in (12).
(20)
/on/ Rhyme-Harmony Id-[back]E-M Id-CPl-Coda
[on] *!
> [oŋ] *
[ən] *!
The behavior of the English rimes composed of the centralized, wedge
vowel [ʌ] suggests that it is not salient enough on the crucial [±back] F2
dimension to force a change in the nasal coda. Faithfulness to the coda
obtains in all but one case (21). The vowel receives a range of adaptations
as high, mid, or low.
(21) English Mandarin
a. uncial [ʌn] an.se.er [an]
27
punch pan.qu
hundredweight han.jue.huai.tuo
carborundum ka.bo.lan.deng
Brunswick bu.lun.si.wei.ke [un]
sundae sheng.dai [ɤŋ]
b. Young [ʌŋ] yang [ɑŋ]
To summarize the adaptation of English VN rimes with a mid vowel,
we find [ɛn] and [on] primarily rendered as faithful to vowel quality at
the expense of change in the nasal consonant. The adaptation of [ʌn] and
[ʌŋ] is determined by the nasal coda, indicating that the wedge vowel is
not decisive and reflecting its intermediate position on the [±back] F2
dimension.
Before turning to reduced vowels, we note a minor pattern. Seven
items in our corpus terminate in the graph -oon. Since [dun], [tun], [sun]
and [lun] are valid Mandarin syllables, it is puzzling why these
adaptations are rejected in five of the seven items--primarily in favor of
C[ong]. Possibly in these cases the adapter was following a graphic rule
28
that interprets the -oon as if it terminated in a tense mid vowel, (i.e. [on])
rather than the phonetic [un].
(22) English Mandarin
cardoon [un] ha.dun [un]
monsoon mang.xun [yn]
cartoon ka.tong [oŋ]
pantaloon pa.ta.long
pontoon pang.tong
shalloon xia.long
simoon xi.meng [ɤŋ]
English loans with final syllables containing unstressed, reduced
vowels exhibit two competing adaptation strategies. The primary one
substitutes en [ən]--arguably Mandarin's best phonetic match to the
schwa-like, reduced vowel of English (23a). This practice is followed
unless an illegal or rare syllable such as len [lən] or den [tən] results, in
which case a high vowel is typically substituted (23b) instead. In a few
cases (23c), the adapter has based the choice on the spelling.
29
(23) English Mandarin
a. Addis [ən] a.di.sen [ən]
eikonogen ai.ke.nu.zhen
Bremen bu.le.men
predicament bu.li.di.jia.men
cushion gu.chen
claisen ke.lai.sen
co.se.cant kou.xi.gen
li.nen lian.ren
mammon ma.men
Mormon mo.men
bacon pei.gen
pullman pu.er.men
salmon sa.men
cinchophen xin.ke.fen
union yu.ren
b. Appleton [ən] a.pu.dun [un] [dən] (rare)
dal.ton dao.er.dun
30
weston wei.si.dun
baron ba.lun *lən
per.lon bei.lun
gallon jia.lun
carron ka.lun
Corbillon kao.bi.lun
chaldron qiao.te.lun
c. satan [ən] sa.dan [an]
ti.tan tai.tan
Zion xi.an
harmattan ha.ma.dan
These data can be analyzed by assuming that the schwa of the source
word is not salient enough to determine the outcome and the decision is
passed on to the coda nasal. Since all the examples have coda [n], no
change is required. Our corpus contains no unstressed syllables with a
coda dorsal nasal, which in any case are rare with nonhigh vowels in
English. The tableau below illustrates the adaptation baron > ba.lun. The
adaptations with the high vowel in (23b) indicate that [+high] is
31
preferred to [+low] as a match for the unstressed schwa of English,
probably because high vowels are phonetically shorter than low vowels
cross-linguistically.
(24)
/bærən/ Use-Ld-Syll Id-[back] Id-CPl-Coda Id-duration
lən *!
loŋ * *!
lan * *!
>lun *
In sum, the adaptations analyzed in this section indicate that when the
vowel of the English source word is front or back then it determines the
way in which the loan accommodates the Rhyme Harmony constraint.
Nonsalient schwa or wedge seem to pass the decision on to the nasal
coda.
In the next two sections we review a couple of other places in the
loanword grammar where the place feature of a nasal coda is determined
by the vowel of the source word
32
2.2 V.NV -> VN.NV
In (25) we list examples in which a nasal consonant is added to the coda
before a following nasal onset in order to satisfy the bimoraic
requirement on stressed syllables. Interestingly, the choice between [n]
and [ŋ] is determined, not by geminating the nasal of the source word,
but rather by the vowel of the augmented syllable (25a). For example, in
the adaptation of economy the English stressed syllable is augmented in
the Mandarin loan by insertion of a velar rather than a dental nasal:
ai.kang.nuo.mi. The handful of exceptions to this generalization is shown
in (25b).8
(25) English Mandarin
a. amonal [on] a.mang.na [ɑŋ]
economy [ɑn] ai.kang.nuo.mi [ɑŋ]
anarchy [æn] an.na.qi [an]
8 The OED indicates a back vowel for the medial syllable of Afghani in (25b). For this reason we classify it as an exception. Also the loan mammoth > meng.ma is represented with the character for meng 'fierce', perhaps for semantic reasons.
33
benadryl [en] ben.na.jun [en]
felony [ən] fei.lun.nu [un]
laudanidine [æn] lan.dan.ni [an]
mana [æn] man.na [an]
monarchy [ɑn] meng.ne.'a.ji [ɤŋ]
perphenazine [en] pia.fen.na.xin [en]
penicilin [en] pan.ni.xi.lin [an]
thiram [æm] qiu.lan.mu [an]
seneca [en] sen.ni.jia [en]
penny [en] pen.ni [en]
arsphenamine [en] shen.fan.na.ming [an]
spanner [æn] shi.ban.na [an]
scammony[æn] si.kan.mo.ni [an]
Tammany [æm] tan.mu.ni [an]
gunny [ʌn] gong.ni [oŋ]
Tony [on] tang.ni [ɑŋ]
Downing [ɑun] tang.ning [ɑŋ]
b. afghani [ɑn] a.fu.han.ni [an]
memory [em] meng.mo.li [ɤŋ]
34
mammoth [æm] meng.ma [ɤŋ]
The data in (25) show that faithfulness to the backness of the vowel--
redundant in Mandarin but contrastive in English--is an active constraint
of the loanword grammar that overrides homorganicity for the NC cluster
that might otherwise be expected since it does not require the insertion of
a place feature in the coda but merely anticipates the place feature of the
following onset. We illustrate this aspect of the Mandarin loanword
grammar in (26).
(26)
e/kɑn/omy Id-V[back]E-M Dep-Place-Coda
ai.kan.nuo.mi *!
>ai.kɑŋ.nuo.mi *
2.3 Vm -> Vn, Vŋ
Finally, in (27) we list loans where the English source word contains a
labial nasal in the coda. Since Mandarin bars [m] from the coda, the nasal
35
coda must alter its place of articulation. The data indicate that the choice
between [n] and [ŋ] is determined primarily by the front vs. back nature
of the preceding vowel in the English source word (27a). The more
centralized wedge vowel is once again less decisive, occurring with both
dorsal and well as coronal nasal codas. Exceptions are shown in (27b).9
Here as well we find that the adaptation has recourse to the more salient
vowel rather than substituting a default consonant such as [n] that might
otherwise be expected under a *Dorsal » *Labial » *Coronal ranking for
consonantal place (de Lacy 2006).
(27) English Mandarin
a. ambersite [æm] an.bu.rui.te [an]
ambroise [æm] an.bu.luo.si
ampul [æm] an.bu
samsonite [æm] san.suo.na.te
Gram [æm] ge.lan
jam [æm] zhan
compost [ɑm] kang.po.si.te [ɑŋ] 9 The shampoo > xiang.bo loan is represented with the character for 'fragrance' and so may be a case of semantic contamination.
36
combination [ɑm] kang.bai.na.xiong
compost [ɑm] kang.po.si.te
compote [ɑm] kang.bo.te
communism [ɑm] kang.men.ni.si.mu
commons [ɑm] kang.men.si
combiner [ɑm] kang.ping.na
commission [ɑm] kang.mi.xiong
Tom [ɑm] tang.mu
Thompson [ɑm] tang.mu.sheng
samba [ɑm] sang.ba
quinoform [om] kui.nuo.fang
embelin [em] en.bei.lin [ən]
sumbul [ʌm] sang.bo [ɑŋ]
gumbo [ʌm] gong.bo [oŋ]
rum [ʌm] lan.mu [an]
calumba [ʌm] ka.lun.ba [un]
yumpies [ʌm] yong.pi.si [oŋ]
carborundum [ʌm] ka.bo.lan.deng [ɤŋ]
trumpet [ʌm] qu.lang.pai.ti [ɑɳ]
37
atom [əm] a.tun [un]
Edam [əm] yi.dun [un]
b. shampoo [æm] xiang.bo [ɑŋ]
mambo [ɑm] man.bo [an]
adam [əm] ya.dang [ɑŋ]
empire [em] ying.bai.er [iŋ]
emperor [em] ying.bai.li.re.er [iŋ]
The tableau in (28) shows the effect of Ident-V[back]E-M in the adaptation
of compost.
(28)
/kɑmpost/ Id-V[back]E-M *dorsal, labial *coronal
kan.po.si.te *!
> kɑŋ.po.si.te *
The adaptations of the unstressed syllables of atom and Edam ([atun]
and [yidun]) with a coronal support the idea that [n] is the default nasal.
If the [±back] quality of the schwa vowel of English is indeterminant (as
38
seems plausible--cf. Flemming & Johnson 2007) then the choice between
the coronal and dorsal coda is resolved by the markedness hierarchy that
substitutes coronal as the default oral place.
(29)
/ætəm/ Id-V[back]E-M *dorsal *coronal
>atun *
atuŋ *!
Before concluding this section we briefly address the possible role of
orthography in the coda nasal adaptation process. The vast majority of
loans spelled with "on" are adapted as [ɑŋ] and those spelled with "an" as
[an]. Could orthography be the basis of the adaptation pattern rather
than reference to the salience of the vowel on the F2 dimension? We
think not. First, the few words in our corpus with an [ɑn] sequence from
Romance languages such as franc and canto are adapted with a back
rhyme in accord with the back vowel in the source. Second, "on" and "an"
words where the corresponding syllable in English is unstressed such as
cushion are by and large adapted with [ən] and not [ɑŋ] or [oŋ]. Since
39
stress is not orthographically recorded, the adaptation must be based on
the spoken form of the word to explain this distinction. In our view the
adapters use their knowledge of the spelling regularities of the source
languages to guide them in the correct pronunciation of the source word
vowel, which in turn determines the adaptation. This is evident in
occasional mistaken interpretations such as satan > [sa.dan] where the
final syllable is treated as stressed. Finally, even if we were to grant that
orthography is the basis of the adaptation, it does not help to explain why
in the orthographic equivalence of "on" = [ɑŋ], it is the vowel symbol
that is the determining factor rather than the consonantal one. Salience in
the phonetics provides a more plausible basis for understanding the
adaptation, especially when it is combined with the observation that the
less salient central vowels wedge and schwa do not determine the
outcome. Here it is the nasal consonant that appears to do so. If the
Mandarin adaptation of nasal codas is based on spelling, then if "n"
determines the outcome for syllables with wedge and schwa, why not for
"an" and "on" as well?
40
3. Another Corpus
The list of place names on the Chinese Ministry of Foreign Affairs website
provides another opportunity to study the adaptation of nasals into
Mandarin.10 These data largely corroborate the generalizations found in
the data from the dictionary discussed in section 2. First, source words
with a front [æ] followed by a nasal ([m] or [n]) uniformly have the [an]
correspondence in Mandarin.
(30) English Mandarin
Alexander [æn] ya.li.shan.da [an]
Amsterdam a.mu.si.te.dan
Anatolia an.na.tuo.li.ya
Atlanta ya.te.lan.da
Birmingham bo.ming.han
Canberra kan.pei.la
Canterbury kan.te.bo.lei
Canton kan.dun
Fanning fan.ning 10 We thank Ross Foo for providing us with a transcribed list of such words.
41
Flanders fo.lan.de
Grampian ge.lan.ping
Hampshire han.pu
Indiana yin.di.an.na
Kansas kan.sai.shi
Manchester man.che.si.te
Manhattan man.ha.dun
Mansfield man.si.fe.ier.de
Nancy nan.xi
Nantucket nan.ta.ji.te
Nottingham nuo.ding.han
Stamford si.tan.fu
In cases where the loan source contains a conflicting combination of
vowel nucleus and nasal coda, the vowel is the determining factor in the
adaptation in the vast majority of cases (31). For [ɑn] Ontario and Tucson
(31b) are the exceptions where we find an unexpected front vowel.
Perhaps the latter is based on a false parsing Tuc+son (cf. Addison >
42
adisen). For [æŋ] the only exception is Doncaster (31d), for which the
OED provides a [dæn] transcription despite the spelling.
(31) English Mandarin
a. Adirondacks [ɑn] a.di.lang.da.ke [ɑŋ]
Bronx bu.lang.ke.si
Connacht kang.nuo.te
Cornwall kang.woer
Klondike ke.lang.dai.ke
Oregon e.le.gang
Wisconsin wei.si.kang.xing
Longford lang.fu.de
Taunton tang.dun
Tyrone di.long [oŋ]
Yukon yu.kong
Montana meng.da.na [ɤŋ]
Monte Carlo meng.te.ka.luo
Montpelier meng.bi.liai
Vermont fo.meng.te
43
b. Ontario [ɑn] an.da.lue [an]
Tucson tu.sen [ən]
Pondicherryu ben.di.zhi.li
c. Anchorage [æŋ] an.ke.lei.qi [ an]
Anguilla an.gui.la
Angus an.ge.si
Frankfurt fa.lan.ke.fu
Franklin fu.lan.ke.lin
Lancashire lan.kai.xia
d. Doncaster [æŋ] tang.ke.si.te [ɑŋ]
We have seven examples for [ʌŋ]. Five are faithful to the nasal,
recapitulating the behavior seen earlier in (21). The syllable gaps *len and
*den motivate the changes in vowel height.
(32) English Mandarin
Fundy [ʌn] fen.di [en]
Brunswick bu.lun.rui.ke [un] *len
London lun.dun *len, *den
44
Dunkirk dun.ke.er.ke *den
Front fu.lan.te [an] *len
Sunderland sang.de.lan [ɑŋ]
Dundee deng.di [ɤŋ] *den, *din
Finally, when the final syllable in the English loan is unstressed, the
expected [Cən] is found in many cases (33a). (33b) and (33c) reflect two
alternative responses to the Mandarin syllable gaps against the otherwise
expected [ən] rime. In the former the vowel is adapted as high (Id-CPl-
Coda » Id-V[high]) while in the latter the height change is blocked,
compelling a change of the nasal coda (Id-V[high] » Id-CPl-Coda). The
adaptations in (27d) appear to be based on the orthographic
representations in which the vowels are treated as full rather than
reduced.
(33) English Mandarin
a. Cardigan(shire) [ən] ka.di.gen [ən]
Devon de.wen
Lincoln(shire) lin.ken
45
Logan luo.gen
New Haven niu.hei. wen
Saxony sa.kr.sen
Solomon suo.luo.men
b. Boston [ən] bo.shi.dun [un] *den
Eton yi.dun *den
Lachlan la.ke.lun *len
Lawrence lao.lun.si *len
c. Croyden ke.luo.yi.deng [ɤŋ] *den
Wimbledon wen.bu.er.deng *den
d. Akron [ən] a.ke.long [oŋ]
Cro-Magnon ke.lu.mai.nong
Birmingham bo.ming.han [an]
Evans ai.fan
Michigan mi.zhi.an
New Orleans niu.ao.er.liang [ɑŋ]
In sum, the adaptations in the place names largely conform to the
generalizations uncovered in the analysis of the dictionary loans in
46
section 2.11 The adaptation of the coda nasal is determined by the position
of the vowel in the source word on the front-back, F2 dimension. When
the vowel occupies an intermediate position on this dimension, as in the
case of wedge [ʌ], or is indeterminant, as in the case of schwa, the nasal
place of the coda is largely preserved.
4. Phonetic Basis
The surface phonetic contrast in vowels before the alveolar vs. dorsal
codas has been studied in a number of phonetic investigations of
Mandarin. For example, Chen (2000) reports F2 differences of c. 500 Hz.
in [an] vs. [aŋ] rhymes when they appear before a stop such as daŋ.da
‘single hit’ (tennis) vs. fɑŋg.da ‘magnify’. They are located in the interior
of the vowel and are not just a coarticulatory effect at the VC transition.
Crucially, she also finds that these differences persist--at a lower (c. 250
Hz) but still significant (P < 0.001) magnitude--in the wake of the
11 The appendix to Hall-Lew's (2002) study of more recent western loanwords into Mandarin drawn from the area of popular culture contains c. 130 items including the following that conform to the generalizations seen in sections 2 and 3: carnation > kang.nai.xin, champagne > xiang.bin, crayon > gu.li.rong, hamburger > han.bao, nylon > ni.long, sandwich > san.ming.zhi, and sauna > sang.na.
47
deletion of an intervocalic nasal in casual speech sha(n).ao ‘cove’. The
magnitude of the F2 differences in Mandarin [an] vs. [ɑŋ] rhymes was
further documented by Mou (2006), who found a c. 400 Hz difference at
the mid point of the vowel for her Beijing subjects (see Figure 1).
Figure 1. Averaged values of F2 movement for 18 Standard Chinese
(Mandarin) vowels from 100 ms. prior to the nasal consonant to 30 ms.
into the nasal consonant
Another relevant finding by Mou (2006) was that the average F2 values
for predorsal low and mid vowels are relatively close to the values found
48
in syllables lacking a coda while the pre-coronal nuclei are more distant
from such open syllables.
(33) F2 in Hz
Ca 1111 Ce 1440
Caŋ 1172 Ceŋ 1448
Can 1330 Cen 1578
This difference makes sense under Flemming's (2003) interpretation of
the relation between coronal consonants and vocalic tongue body features
as one of fronting the tongue body to accommodate a consonantal
constriction at the alveolar ridge. The relatively steady rise in F2 for [an]
in Figure 1 in comparsion to the largely flat trajectory in [ɑŋ] also makes
sense in these terms. Finally, Mou (2006) reports gating experiments in
which her subjects could reliably guess the presence and identity of the
upcoming coda nasal when they heard less than half of a low or mid
vowel. On the other hand with high vowels, where there is a contrast
among [i], [y], and [u], speakers could not reliably identify the following
nasal--especially after [i] where there may even be neutralization of the
49
[n] vs. [ŋ] contrast. In sum, Rhyme Harmony is a genuine process of
Mandarin grammar--an enhancement effect (Keyser & Stevens 2006) that
speakers can utilize to identify the place of articulation of the nasal coda.
5. Summary and Conclusion
This study utilized the Mandarin nasal codas to probe the phonological
vs. phonetic bases of loanword adaptation. Nonhigh vowels are assigned
different allophones along the front--back dimension in order to enhance
a phonemic contrast between coronal and dorsal nasal codas. Our
principal finding is that when the adapter is presented with conflicting
choices to satisfy this phonotactic constraint of native grammar, it is the
information found in the phonetically more salient vowel that determines
the outcome. This result is in line with other cases of such conflict in
loanword adaptation reported in Kenstowicz (2003). Coupled with the
observation that stressed syllables are often the site of cyclic transfer
(Kenstowicz 1997, Steriade 1999), it suggests that perceptual salience
constitutes an alternative dimension of phonological faithfulness.
Tasks for future research include more extensive documentation and analysis of current loanword adaptation patterns in Mandarin as well as a
50
more quantitative analysis of the Rhyme Harmony process along the lines of Flemming (2008). More generally, our study raises the question of whether enhancements which play a role in speech perception couple together features or cues that have a natural, cross-linguistically recurrent relation such F0 and duration cues to consonantal laryngeal contrasts (Hsieh & Kenstowicz 2008); or can they involve more phonetically arbitrary connections that are rooted in the accidents of the history of individual languages? This is of course a fundamental question that has emerged in the field of phonology more generally in the past decade. We believe that continued study of loanword adaptation may provide crucial evidence to help resolve this matter.
References
Bauer, Robert. 1985. “The expanding syllabary in Hong Kong Cantonese”.
Cahiers de linguistique Asie orientale 14.99-111.
Chen, Marilyn. 2000. “Acoustic Analysis of Simple Vowels Preceding a
Nasal in Standard Chinese”. Journal of Phonetics 28.43-67.
De Lacy, Paul. 2006. Markedness: Reduction and Preservation in Phonology.
(=Cambridge Studies in Linguistics 112.) Cambridge: Cambridge
University Press.
Duanmu, San. 2000. The Phonology of Standard Chinese. Oxford: Oxford
University Press. (2007, 2nd edition.)
51
Flemming, Edward. 2003. “The Relation Between Coronal Place and
Vowel Backness”. Phonology 20.335-73.
Flemming, Edward & Sarah Johnson. 2007. “Rosa’s Roses: Reduced
Vowels in American English”. Journal of the International Phonetic
Association 37.83-96.
Flemming, Edward. 2008. “The Grammar of Coarticulation”. To appear in
La Coarticulation: Indices, Direction et Representation, ed. by M. Embarki
& C. Dodane.
Hall-Lew, Lauren Asia. 2002. English Loanwords in Mandarin Chinese. BA
Honors Thesis, University of Arizona, Tucson.
Hsieh, Feng-fan & Michael Kenstowicz. 2008. “Phonetic Knowledge in
Tonal Adaptation: Mandarin and English Loanwords in Lhasa Tibetan”.
Journal of East Asian Linguistics 17.279-97.
Kenstowicz, Michael. 1997. “Base-identity and Uniform Exponence:
Alternatives to Cyclicity”. Current Trends in Phonology: Models and
Methods ed. by Jacques Durand & Bernard Laks, vol. 1, 363-94. Salford:
University of Salford.
Kenstowicz, Michael. 2003. “Salience and Similarity in Loanword
Adaptation: a Case Study from Fijian”. Language Sciences 29.316-40.
52
Kenstowicz, Michael. 2005. “The Phonetics and Phonology of Korean
Loanword Adaptation”. Proceedings of the First European Conference on
Korean Linguistics ed. by S-J. Rhee, 17-32. Seoul: Hankook Publishing
Co.
Kenstowicz, Michael & Atiwong Suchato. 2006. “Issues in Loanword
Adaptation: a Case Study from Thai”. Lingua 116.921-49.
Keyser, Jay Samuel & Kenneth Noble Stevens. 2006. “Enhancement and
Overlap in the Speech Chain”. Language 82.33-63.
LaCharité, Darlene & Carole Paradis. 2005. “Category Preservation and
Proximity vs. Phonetic Approximation in Loanword Adaptation”.
Linguistic Inquiry 36. 223-58.
Liu Zhengtan, Gao Mingkai, Mai Yongqian, & Shi Yowei. 1984. Dictionary
of Loanwords and Hybrid Words in Chinese. Shanghai cishu chubanshe,
Shanghai.
Mou, Xiaomin. 2006. Nasal Codas in Standard Chinese: a Study in the
Distinctive Feature Theory. Ph.D. dissertation, Massachusetts Institute of
Technology.
53
Novotná, Zdenka. s1967. “Linguistic Factors of the Low Adaptability of
Loan-words to the Lexical System of Modern Chinese”. Monumenta
Serica 26.103-18.
Paradis, Carole & Darlene LaCharité. 1997. “Preservation and Minimality
in Loanword Adaptation”. Journal of Linguistics 33.379-430.
Silverman, Daniel. 1992. “Multiple Scansions in Loanword Phonology:
Evidence from Cantonese”. Phonology 9.289-328.
Steriade, Donca. 2000. “Paradigm Uniformity and the Phonetics-
Phonology Boundary”. Acquisition and the Lexicon ed. by Michael Broe
& Janet Pierrehumbert (=Papers in Laboratory Phonology 5.), 313-34.
Cambridge: Cambridge University Press.
Wang, Jenny Zhijie. 1993. The Geometry of Segmental Features in Beijing
Mandarin. Ph.D. dissertation, University of Delaware.
Yip, Moira. 2006. “The Symbiosis Between Perception and Grammar in
Loanword Phonology.” Lingua 116.950-75.